Hello! I’m Inpyo Song, a Ph.D. student researching computer vision and artificial intelligence, advised by Prof. Jangwon Lee. I also collaborate with Prof. David Crandall from Indiana University Bloomington.

My research focuses on video understanding, using object-centric representations and vision-language models to recognize and explain human activities and unusual events. I also study reliable evaluation: whether a model’s scores reflect its ability to recognize, explain, and locate events.

You can find my CV here. I am always open to any form of collaboration. If you have any ideas for potential collaboration, or just feel like having a casual chat, please feel free to reach out!

🔥 News

• Jun. 2026: 🎉 “Bounding-Box Trajectories Matter for Video Anomaly Detection” accepted at ECCV 2026.
• Jun. 2026: 🏆 Received the Outstanding Poster Award at the KIBME Summer Conference 2026.
• Aug. 2025: 🏆 Received the Best Graduate Research Award at the 2025 Digital Innovation Talent Symposium.
• May 2025: 🎉 Two papers accepted at ICIP 2025.
• Feb. 2025: 🏆 Selected for I-Corps Korea Program.
• Dec. 2024: 🎉 One paper accepted at WACV 2025.
• Oct. 2024: 🎉 SFTrack presented as Long Oral Presentation at IROS 2024.

1. Frame-Level Evaluation in Weakly Supervised Video Anomaly Detection Mostly Measures Video-Level Ranking

   R

   Inpyo Song, Jangwon Lee

   Under review | [arXiv]

2. A VLM Answer Is Not an Anomaly Score: Rank Compression Across Image and Video Anomaly Detection

   V R

   Inpyo Song, Jangwon Lee

   Under review | [arXiv]

3. Rethinking Open-World Video Anomaly Detection: Diagnosing Definition Blindness

   V R

   Inpyo Song, Jangwon Lee

   Under review | [arXiv]

4. Bounding-Box Trajectories Matter for Video Anomaly Detection

   O

   Inpyo Song, Jangwon Lee

   ECCV'26 | [Project page] | [Paper] | [arXiv] | [Code]

5. Instance-Aligned Captions for Explainable Video Anomaly Detection

   O V R

   Inpyo Song, Eunji Jeon, Joonhyung Kwon, Minjun Joo, Jangwon Lee

   Under review | [arXiv]

6. Real-Time Traffic Accident Anticipation with Feature Reuse

   O

   Inpyo Song, Jangwon Lee

   ICIP'25 | [Project page] | [Paper] | [arXiv] | [Code]

7. Anomaly Detection for People with Visual Impairments Using an Egocentric 360-Degree Camera

   R

   Inpyo Song, Sanghyeon Lee, Minjun Joo, Jangwon Lee

   WACV'25 | [Project page] | [Paper] | [arXiv] | [Code]

8. SFTrack: A Robust Scale and Motion Adaptive Algorithm for Tracking Small and Fast Moving Objects

   O

   Inpyo Song, Jangwon Lee

   IROS'24 Long Oral | [Project page] | [Paper] | [arXiv] | [Code]

9. Action-Conditioned Contrastive Learning for 3D Human Pose and Shape Estimation in Videos

   O

   Inpyo Song, Moonwook Ryu, Jangwon Lee

   CVIU'24 (IF 4.3, Q1) | [Paper] | [Code]

10. Motion-Aware Heatmap Regression for Human Pose Estimation in Videos

    O

    Inpyo Song, Jongmin Lee, Moonwook Ryu, Jangwon Lee

    IJCAI'24 (14.0% acceptance rate) | [Paper]

11. Video Question Answering for People with Visual Impairments Using an Egocentric 360-Degree Camera

    V

    Inpyo Song, Minjun Joo, Joonhyung Kwon, Jangwon Lee

    CVPRW'24 (First Joint Egocentric Vision (EgoVis) Workshop) | [arXiv]

Select at least one topic to show publications.

See all 20 publications →

🎖 Honors and Awards

• Jun. 2026 Outstanding Poster Award, KIBME Summer Conference
• Aug. 2025 Best Graduate Research Award, 2025 Digital Innovation Talent Symposium, Ministry of Science and ICT
• Feb. 2025 Selected for I-Corps Korea Program (VLM-based Arduino Tutor)
• Oct. 2024 Excellence Award, SKKU Graduate Student Startup Competition (Multimodal VQA-Based Tutoring System for the Digital Transformation of Hands-On Education)
• Aug. 2024 President’s Award from IITP, 2024 Digital Innovation Talent Symposium (Motion-Aware Heatmap Regression for Human Pose Estimation in Videos)
• Feb. 2024 Encouragement Award, SKKU Research Matters (Technological Social Responsibility: Anomaly Detection for People with Visual Impairments Using an Egocentric 360-Degree Camera)
• Dec. 2022 Grand Award from SKKU, KAU Start-up Idea Competition (Generating Digital Twin using Instance-NeRF)

💻 Internships

• Jun. 2024 - Aug. 2024, Research Intern @ Indiana University Bloomington CVLab, Bloomington, IN, USA

🤝 Professional Services

Conference Reviewer

• AAAI 2027
• NeurIPS 2026
• CVPR 2026
• IROS 2026
• AAAI 2026
• WACV 2025
• ICDL 2024